Sometimes the workflow manager crashes during tasks.

[acreegan]

Normally this is prevented since every task runs in a try-except block.
I suspect the crashes are caused by segmentation faults in some meshing libraries.
This needs to be further investigated.

One solution may be to run tasks in a separate process even when use_multiprocessing is not activated.

[acreegan]

With pool.imap, it hangs if a child process has a segfault. See python/cpython#66587. But maybe we can solve this with another method for creating subprocesses? Concurrent.futures? https://stackoverflow.com/questions/24370930/multiprocessing-pool-hangs-if-child-causes-a-segmentation-fault